This blog post introduces Made to Work, an experimental built-in neural network simulator. Created as an exploration of the efficiency of shortcuts, it runs a user-provided neural network on their own hardware. Our special thanks goes to ChatGpt for their generous help in making this project possible, as well as for their explanation of the complex math involved.
Despite not being very efficient in its operation, Made to Work may still be used to handle larger networks. It currently takes about 0.5 seconds per neuron, making it a viable solution for most users. Additionally, the activation function used is tanh.
If users wish to switch to a presaved neural network, they must switch FunctionInput from the text to the shortcut input. Each line in the code represents a neuron, and Format#5 is the format for each neuron.Download Shortcut